Data Integration. PDI handles pretty much any type of data source and can transform data and conform it to any type of business logic.
Object Oriented ETl Coding. Code developed in PDI can be shared across numerous jobs/transformations. This allows for resusable and maintainable code.
Flexibility and plugins. If the function that you are looking for is not built into the tool, it's fairly straightforward to either download or develop plugins. There is a huge community of users that build new functionality all the time.
Scaleable and cloud ready. PDI is able to cluster and scale out to handle huge data sets.
Talend and Pentaho have a lot of the same functionality, but Talend's interface is not as intuitive. Talend generates code that is then executed while Pentaho is an engine based tool with highly optimized Java code templates that are compiled at runtime.
PDI is a great alternative to proprietary tools. If your environment is just starting out using ETL tools, definitely look into using PDI. The number of features plus the ability to write flexible dynamic processes allows for much more maintainable code.
1 - Knowledge of Business Intelligence, Data Integration, and data usage is a must. The tools are relatively easy to learn with the given references, community, and general ecosystem but it does take time to master - as with any tool set.